agentleFS
ChaiWithJaiChaiWithJai/davinci-resolve-claude-skills/skills/davinci-resolve-cut-screen-recording/SKILL.md

davinci-resolve-cut-screen-recording

Skill19 starsChanged 6 months agoMITView on GitHub

Use when a user has a long screen recording, tutorial, conference talk, or demo recording and needs to cut it down to a tight watchable version — removes dead air, mistakes, "ums", and silences. Triggering symptoms include phrases like "raw recording is too long", "edit out the bad takes", "fix the ums", "remove silences", "rough cut my demo", or any DaVinci Resolve cutting question for long-form content.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 8 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When NOT to use:

    line 21
  2. Distinctive:

    Wait — about 1 minute per 10 minutes of audio.

    line 42
  3. Distinctive:

    If you have free Resolve (not Studio), AI Transcription is unavailable.

    line 44
  4. Very specific:

    Create a new bin called TIMELINES and a new timeline inside it called ROUGH_CUT.

    line 48
  5. Very specific:

    Use J / K / L to scrub backward/stop/forward through the transcript.

    line 50
  6. Distinctive:

    Search the transcript for keywords — type a word in the Search field at the top of the Transcription window to jump to every instance.

    line 55
  7. Distinctive:

    Use Shift-, / Shift-. to nudge an edit by 5 frames forward/back.

    line 75
  8. Distinctive:

    When you hear a mistake / dead air / re-take, press I for In-point.

    line 82
  9. Distinctive:

    Every "um" / "uh" / long pause has been removed or trimmed.

    line 117
  10. Distinctive:

    If the user reports that transcription edits land on the wrong word, breaths are cut weirdly, or Remove Silent Portions misses pauses, read …

    line 133

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

DaVinci Resolve — cut a long screen recording into a tight demo

Overview

The fastest way to turn a 60-minute screen recording into an 8-minute watchable demo. Uses Resolve's AI Transcription plus Remove Silent Portions to do 80% of the work automatically, then refines manually. The core principle: do not waste your time scrubbing a waveform when Resolve can find the dead air for you.

When to use

Symptoms:

  • You recorded a 30-90 minute demo, talk, or tutorial and need to ship a tight version
  • You said "um" too many times
  • There are long pauses between sentences where you were thinking
  • You did 2-3 takes of the same explanation and need to pick the best one
  • Phrases like "raw recording is too long", "remove silences", "fix the ums"

When NOT to use:

  • Multi-camera interview / talk-show — that needs the Multicam workflow (Editor's Guide ch. 4).
  • Tight musical edits to a beat (rhythmic cuts) — use the Cut page manually.
  • The recording is already under 10 minutes — manual editing is faster than this skill.

Quick reference

ToolWhat it doesWhere
AI TranscriptionGenerates word-level transcript with speaker detectionRight-click clip > Audio Transcription > Transcribe With Speaker Detection
Remove Silent PortionsAuto-deletes silent ranges from the timelineTranscription window > Options menu (...) > Remove Silent Portions
Editing from transcriptClick a word, press F12, the clip lands in the timeline at that wordTranscription window > Append (F12) / Insert (Shift-F12) buttons
Subframe audio editingEdit audio at sub-frame precision for tight cutsCmd-B for Razor, then trim by mouse

Steps

Phase 1 — Set up and transcribe (3 minutes)

  1. Switch to the Edit page (Shift-4).
  2. In the Media Pool, drag your screen recording into a bin (e.g. 01 — Footage).
  3. Right-click the clip > Audio Transcription > Transcribe With Speaker Detection. (This is Studio-only; if you have free Resolve, use the manual Cut workflow below.)
  4. Wait — about 1 minute per 10 minutes of audio. The AI Transcription window opens automatically when done.

If you have free Resolve (not Studio), AI Transcription is unavailable. Use the Cut page approach: import the clip, drag to the timeline, use the mouse-scrub-and-mark-In/Out workflow on the Cut page (Beginner's Guide pp. 1-67 covers this).

Phase 2 — Create the rough cut from the transcript (10-15 minutes)

  1. Create a new bin called TIMELINES and a new timeline inside it called ROUGH_CUT. Audio Track Type = Mono if your recording is mono voice.
  2. With the Transcription window open, click on any word — the playhead jumps to that timecode in the source viewer.
  3. Use J / K / L to scrub backward/stop/forward through the transcript.
  4. To add a soundbite to the timeline:
    • Click and drag through the words you want to keep — this sets In and Out points automatically.
    • Press Shift-F12 to Append the marked range to the timeline end.
    • Or press F12 to Insert at the playhead.
  5. Search the transcript for keywords — type a word in the Search field at the top of the Transcription window to jump to every instance.
  6. Build the rough cut by appending each good soundbite one by one. Aim for ~2x your target length (a 16-minute rough cut for an 8-minute final).

Editor's Guide pp. 352-358 covers this workflow in depth.

Phase 3 — Auto-remove silent portions (1 minute)

  1. In the Transcription window, click the Timeline button (top-left of the Transcription window) to switch from clip-mode to timeline-mode. The window now shows the transcript for everything on the timeline, with silent gaps marked as (...).
  2. Click the Options menu (three dots ...) > Remove Silent Portions.
  3. The silent ranges are ripple-deleted from the timeline. Your 16-minute rough cut is suddenly 12 minutes.

Editor's Guide p. 358 shows this exact step.

Phase 4 — Refine the edits manually (10-20 minutes)

The AI cuts are crude — they leave breaths cut off or pauses that feel unnatural. Walk the playhead through each edit point.

  1. Press N to disable snapping (so you can trim by mouse at sub-frame precision).
  2. Press Shift-S to enable audio scrubbing (you can hear what you are passing over).
  3. At each cut: hover the cut point, click and drag the edge to extend or pull in by a few frames.
  4. Use Shift-, / Shift-. to nudge an edit by 5 frames forward/back.

The audio subframe editing tip is on Editor's Guide p. 496.

Phase 5 — Cut the obvious mistakes (5 minutes)

  1. Play through end-to-end at 1.5x speed.
  2. When you hear a mistake / dead air / re-take, press I for In-point.
  3. Find the end of the mistake, press O for Out-point.
  4. Press Shift-Delete to ripple-delete the marked region.
  5. Continue. This pass should remove 1-3 minutes more.

You should be at or near your target length. Save the project.

Python automation (rough silence detection on existing timeline)

For a long talk you have already imported, this script walks the active timeline, finds audio clips, and prints their durations so you can quickly identify long silent gaps to manually remove. It is not a substitute for AI Transcription's Remove Silent Portions — it is a helper for when you need to script a multi-clip operation.

The script uses the official DaVinci Resolve scripting API documented in DaVinci Resolve/Developer/Scripting/README.txt and the public Reference Manual.

# scripts/auto_silence_cut.py
# Marks silent regions on the active timeline by inspecting clip start/end frames.
# Run from inside Resolve via Workspace > Console > Py3, or from terminal after
# setting RESOLVE_SCRIPT_API / RESOLVE_SCRIPT_LIB / PYTHONPATH (see README.md).

See scripts/auto_silence_cut.py for the full script. The script is conservative — it does not actually delete anything. It prints clip-level timecode reports so you can verify before running destructive edits.

Common mistakes

  • Trusting the AI Transcription's accuracy -> it gets technical terms, brand names, and acronyms wrong. Always proofread before treating it as a script. Editor's Guide pp. 346-348 documents the Search + Replace workflow for fixing systematic errors (e.g. "Oregon" -> "Organ" misrecognition).
  • Cutting on the Cut page when this is a single long clip -> the Cut page is optimized for fast assembly from many small clips. For one long recording, the transcript-driven workflow on the Edit page is faster.
  • Forgetting to disable snapping when fine-trimming -> snapping forces your cuts to align with existing edits and frame boundaries. Press N to toggle it off for fine work, on for assembly.
  • Doing the entire pass at 1x playback speed -> 1.5x or 2x is faster and your ear can still spot rough cuts. Use the L key twice to enable 2x playback.

Verification

You succeeded if all of the following are true:

  1. The timeline duration matches your target (or is within 30 seconds of it).
  2. Playing back the timeline end-to-end, you cannot hear any cut edits — the audio flows smoothly.
  3. Every "um" / "uh" / long pause has been removed or trimmed.
  4. The audio levels look broadly even on the meters (no clip is much louder than another — though the next skill, davinci-resolve-audio-cleanup-podcast, handles the actual leveling).
  5. You can hand the timeline to a teammate and they cannot tell which cuts were AI-driven vs manual.

Transfer

Now try this on a 20-minute conference talk recording. Watch the silence-detection threshold — large auditoriums often have ambient room noise above -30 dB, so Resolve's Remove Silent Portions might not detect actual pauses. In that case, you may need to manually search for the "(silent)" markers in the transcript and remove them by hand, or apply a quick noise-reduction pass first (see davinci-resolve-audio-cleanup-podcast) so the noise floor drops below the silence-detection threshold.

Working reference

  • docs/wiki/editors-guide.md#lesson-6--ai-workflows-pp-323-387 (AI Transcription, Remove Silent Portions — primary)
  • docs/wiki/beginners-guide.md#lesson-1--editing-a-rough-cut-pp-1-67 (Cut/Edit page basics for free Resolve users)
  • docs/wiki/master.md#reset-matrix--when-the-user-pushes-back-read-this (transcription pitfall rows)

When the agent's work isn't matching expectations (context-rot reset)

If the user reports that transcription edits land on the wrong word, breaths are cut weirdly, or Remove Silent Portions misses pauses, read these PDF page ranges to reset:

  • DaVinci-Resolve-20-Editors-Guide.pdf pp. 341-361 (Lesson 6 — full AI Transcription workflow)
  • DaVinci-Resolve-20-Editors-Guide.pdf pp. 352-358 (Editing Using Transcription — append/insert/F12 semantics)
  • DaVinci-Resolve-20-Editors-Guide.pdf pp. 358-360 (Editing Transcribed Clips in the Timeline — Remove Silent Portions)
  • DaVinci-Resolve-20-Editors-Guide.pdf p. 496 (Subframe Audio Editing)
  • DaVinci-Resolve-20_Beginners-Guide.pdf pp. 1-67 (Cut/Edit page basics — fallback workflow for free Resolve, no AI Transcription)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.